Combustion and Flame 136 (2004) 445–456
www.elsevier.com/locate/jnlabr/cnf
Kinetics and visualization of soot oxidation using
transmission electron microscopy
Heejung Jung,a David B. Kittelson,a and Michael R. Zachariah a,b,∗
a Department of Mechanical Engineering, University of Minnesota, 111 Church Street SE, Minneapolis, MN 55455, USA
b Department of Chemistry, University of Minnesota, 111 Church Street SE, Minneapolis, MN 55455, USA
Received 20 February 2003; received in revised form 9 October 2003; accepted 9 October 2003
Abstract
Visualization of the oxidation of individual soot particles using transmission electron microscopy (TEM) was
performed and quantitative oxidation rates were determined. Soot particles were sampled from an ethylene diffusion flame burner and collected on a TEM grid. Individual particles were then tracked by creating maps of TEM
images at different magnifications, and images were compared before and after the oxidation process in a separate
furnace at low temperatures (< 800 K). Mono-area soot particles were sampled on TEM grids using a differential
mobility analyzer (DMA) and a low-pressure impactor. It was found that the mobility equivalent diameter determined by the DMA is equivalent to the projected area equivalent diameter determined by the TEM for particles
below 150 nm. Groups of mono-area particles were visualized before and after the oxidation process and oxidation
rates were determined from the change in projected area equivalent diameter by image-processing a series of TEM
images. For these low temperatures (< 800 K), an activation energy of 148 kJ/mol was found for the oxidation
of soot particles. The overall oxidation rate was found to be significantly higher than in prior studies, and some
possible reasons for the discrepancy discussed.
 2003 Published by Elsevier Inc. on behalf of The Combustion Institute.
Keywords: Soot; Oxidation; Kinetics; Visualization; TEM; Soot combustion
1. Introduction
With the great emphasis being placed on highefficiency power generation, diesel engines are increasingly being looked upon as the technology of
choice. In turn this has intensified the interest in methods and technologies to reduce soot emissions. It is
unlikely that it will be possible to meet future particle
emission standards without the use of aftertreatment
devices like the diesel particulate filter (DPF). Under
typical operating conditions, diesel exhaust temperatures may be quite low, below about 300 ◦ C. The
performance of DPFs is highly dependent on the rate
at which soot oxidizes and requires a more detailed
* Corresponding author.
E-mail address: [email protected] (M.R. Zachariah).
understanding than is currently available, particularly
at low temperatures. Earlier works on the measurement of low-temperature oxidation included thermo
gravimetric analysis (TGA) [1,2] or flow reactors with
stationary soot beds [3,4]. While TGA is convenient,
it is known to be plagued by heat and mass transfer
effects, which make the data interpretation and quantification of the intrinsic kinetics difficult. In a recent
study we have shown that the reactivity of nanoparticles can be underestimated by as much as 3 orders of
magnitude using TGA [5]. Furthermore these studies
also showed that the reaction constants extracted by
these means were very sensitive to the sample mass
used in the TGA. The use of flow reactors with fixed
beds also requires the use of bulk soot samples and is
therefore presumably also corrupted by similar heat
and mass transfer limitations.
0010-2180/$ – see front matter  2003 Published by Elsevier Inc. on behalf of The Combustion Institute.
doi:10.1016/j.combustflame.2003.10.013
446
H. Jung et al. / Combustion and Flame 136 (2004) 445–456
Fundamentally therefore it would be preferable to
assess the reactivity of a single soot particle to extract
the nascent intrinsic chemical kinetic rates. Ishiguro
et al. [6] investigated the morphology change of soot
during oxidation using an electron microscope and
FTIR spectroscopy. Their investigation focused on
microstructural analysis, such as the behavior of crystallites in soot during oxidation. However, they did
not use this approach to quantify the oxidation rate.
Tanaka [7] also visualized the oxidation of soot particles using transmission electron microscopy (TEM).
He collected diesel soot particles on a thin Al2 O3
plate and tracked their change in size using TEM images. He chose a single agglomerate about 0.5 µm in
size and observed that each of the primary particles
gets smaller as it is oxidized.
The work presented in this paper takes the work
of Tanaka [7] a step further. Soot particles extracted
from a diffusion flame were size segregated by differential mobility and deposited on TEM grids. A group
of approximately 100 particles was tracked to examine the change in morphology and size as a function
of oxidation time and temperature. To extract quantitative rates the images were processed in terms of
projected area equivalent diameter. The results of this
paper are the development of an experimental protocol to measure the oxidation rate of soot particles at
low temperatures (< 800 K) using TEM image analysis.
2. Experimental
2.1. Soot source, size selection, and sampling
Figures 1a and 1b show diagrams of the experimental setups used in the current study. Soot particles
were generated using a Santoro-type diffusion flame
burner [8], using ethylene as the fuel. A 10-in. high
glass shroud is placed on the outer tube to prevent the
flame from being disturbed by outside air currents.
The flame was maintained at 50 mm high at an ethylene fuel flow rate of 85 cm3 min−1 .
The sampling probe used in this study is similar
to one described in our prior work [9]. Soot particles were extracted at the tip of the flame through
a 1-mm orifice in the bottom of an 11-mm diameter sampling probe and diluted by a 20 L min−1 flow
of filtered clean dry air. More detailed information
on this burner and the sampling probe has been previously described [9]. Because soot particles generated in the flame are highly charged with many more
multiply charged particles than an equilibrium charge
distribution at room temperature, the extracted soot
particles were sent through a bipolar diffusion charger
(neutralizer) to establish a known charge distribution. Particles were then passed through a differential
mobility analyzer (DMA). The DMA selects monoarea particles based on electrical mobility, which is
related to the drag and charge on a particle rather
than on absolute particle size. This means larger particles with multiple charges will pass through at the
same voltage setting as the singly charged particles
of the size of interest. For the particle sizes used in
the current study, the majority of charged particles in
an equilibrium charge distribution will carry a single
charge and the minority of multiply charged particles will also pass through the DMA. These monoarea particles were subjected to three possible fates
as illustrated in Figs. 1a–1c. In experiments shown in
Figs. 1a and 1b, we used the lowest stage of a lowpressure impactor [10], after removal of the impaction
plates in the higher states, as an efficient means to
deposit particles on a TEM grid. The initial size distribution of particles collected on the TEM grid was
obtained using TEM image analysis. Next, the particles on the TEM grid were oxidized in a furnace
and subsequently underwent TEM analysis to observe
the change in size. Figure 1b shows an experimental system in which particles were oxidized in free
flight in a heated flow tube of known residence time
and temperature profile. The particles were collected
in the same manner as in Fig. 1a and analyzed using TEM analysis. Finally, Fig. 1c shows a tandemDMA experiment we have previously reported on, in
which the change in size after the flow-tube oxidation was measured with a second differential mobility analyzer [9]. The tandem-DMA system shown in
Fig. 1c is used for high-temperature oxidation studies, but is not amenable for studying slow reactions
due to the long residence times required. On the other
hand the system depicted in Fig. 1a, in which the sample is prepared and then placed with the grid in an
oven, cannot be used at high temperatures at which
the chemistry is fast. In order to provide a check of
consistency between these two approaches, we used
the configuration depicted in Fig. 1b, in which, subsequent to oxidation and deposition, we conducted the
same image analysis procedure as used in experimental system 1a.
2.2. Soot oxidation
Following the deposition of the size-selected soot
on the TEM grids, a portion which has sufficient particles to provide statistically meaningful data was selected. A series of TEM images (Model JEOL 1210)
was taken using a digital CCD camera from low to
high magnification as shown in Fig. 2. This enabled
us to create maps on which the same particles could
be located later on. We found that TEM grids with
H. Jung et al. / Combustion and Flame 136 (2004) 445–456
447
(a)
(b)
(c)
Fig. 1. Schematic diagram of the experimental setups. (a) Setup for TEM sampling. (b) Setup to compare the current study with
the results of Higgins et al. [9,18]. (c) The setup of Higgins et al. [9]. The setup used in Higgins et al. [18] is the same except the
particle source is a diesel engine.
holey or lacy film provided easily identifiable geometric features enabling one to relocate particular
particles. Because the grids were to be removed and
subjected to an elevated-temperature oxidizing environment (air) we could not use the standard carboncoated grids, but rather employed nickel mesh with a
holey silicon oxide film.
Once a map of the grid was made the grid could
then be removed and inserted into a furnace for a
specified time and temperature as shown in Fig. 3.
The mechanics of the experiment shown in Fig. 3
involved loading the TEM grid (3 mm in diameter)
into a very low heat capacity stainless steel foil crucible (thickness of 25 µm and 10 × 10 × 3 mm). The
tube furnace, which has a 300-mm heating zone, contained a 25-mm-diameter alumina outer tube and a
15-mm-diameter inner quartz tube. The purpose of
the second quartz tube was simply for convenience
448
H. Jung et al. / Combustion and Flame 136 (2004) 445–456
Fig. 2. Maps of TEM images to track a field of view.
Fig. 3. Schematic diagram of a furnace.
of allowing the sample to be rapidly inserted and removed. To minimize transient temperature effects, the
unloaded crucible was first thermalized in the flow
oven for 1 h to reach steady-state temperature conditions, after which the quartz tube with the crucible
was partially removed to load the TEM grid containing the particles. Despite the care taken to thermalize
the system and the use of low-heat-capacity structures, a transient temperature profile was still evident
as shown in Fig. 4. The temperature profile shown
here was measured with a 0.05-mm-diameter type K
fine wire thermocouple. While this measurement was
made without the crucible, it should be a fairly accurate representation of the time–temperature history
experienced by the particles. We have estimated the
temperature uncertainty to be less than ±2 K and it is
described in Appendix A.
Soot particles on the TEM grid were oxidized in
heated air for a given time and temperature. Analysis
of the oxidation-exposed grid used the same field of
view as was used prior to oxidation using the procedure illustrated in Fig. 2.
3. Results and discussion
3.1. Visualization of soot oxidation
3.1.1. Tracking individual particles
Figure 5 shows examples of individual particles
during oxidation at different cumulative oxidation
Fig. 4. Temperature profile of a furnace, at the place where
the oxidation of soot particles occurs, as a function of
elapsed time.
times at 450 ◦ C. Initial images of particles (a) and
(e) were taken and those particles were tracked separately. The soot aggregates are composed of primary particles in the 20–30 nm size range. Clearly
the primary particles are shown shrinking during this
process, but we also observe some changes in the aggregate morphology. The latter observation presumably has to do with how the particles find themselves
anchored to the grid, and one might expect differences from what might occur in the gas phase. Most of
the primary particles were observed to shrink evenly
during oxidation. For these particular images an arm
of the aggregate seemed to move during oxidation
(Figs. 5a–5d), whereas the particle in Figs. 5e–5h
H. Jung et al. / Combustion and Flame 136 (2004) 445–456
449
Fig. 5. TEM images during oxidation for individual particles.
shrank in size without significant changes in shape.
Our more gradual evolving particle shape and size
are in contrast to the observations of Heckman [11].
He employed high-temperature heat treatment (partial
graphitization in an inert atmosphere) before carrying out an oxidation experiment. His work found very
little shrinkage, but rather an internal hollowing out
of the particles. He speculated that a more ordered
graphitic layer structure, at the surface of the particles, was stable against oxidation, and Gilot et al.
[12] also reported internal oxidation of carbon black
spherules after 60% burn-off at low temperature. We
observed no selective hollowing of soot particles in
this study, which we believe is related to the absence
of a pregraphitization process [13].
To quantify the results of the oxidation we used
a public domain image program (NIH Image [14]) to
obtain the projected particle area. Using the software,
for example, on the images in Fig. 5, we can determine that particle (e) has a projected area equivalent
diameter of 178 nm and that after oxidation for 20 min
the resulting particle (h) has shrunk to a projected area
equivalent diameter of 108 nm. Our goal is to relate
the projected area change to the surface-specific reaction rate for soot oxidation. Since individual particles
can undergo different projected area reduction rates
due to the facts that a portion of a particle can disappear during oxidation and the oxidation rate can be
different due to differences in chemical composition
for individual particles, we tracked a group of particles rather than individual particles to quantify the
oxidation rate.
3.1.2. Tracking a group of particles
A field of view containing about 100 particles was
chosen for tracking during oxidation. In order to observe such a large number of particles in a single im-
age the magnification used (×4000–×5000) is about
six times lower than that used for tracking individual
particles and is shown in Fig. 6a. Figure 6b shows the
image after oxidation in a furnace for 4 min at a set
temperature of 450 ◦ C. Qualitatively we see that particles are clearly shrinking, but that some particles are
seen to be oxidizing faster than others and in some
cases disappear entirely. These differences in oxidation rate would seem to indicate that even though all
particles were sampled from a single location and size
segregated with a DMA, they may not be chemically
the same. Particles that disappeared were excluded
manually during image analysis so that the exact same
group of particles could be tracked.
Figure 7 shows the same area after image processing with the image program (NIH Image [14]). The
image shown is after digital filtering to remove the
background and conversion to a binary image. Pixels
that could not be removed, but that obviously were not
particles, were erased manually. Next, the image program (NIH Image [14]) obtained the projected area
for each particle, from which we could determine the
projected area equivalent diameter (the diameter of a
sphere that has the same projected area as that of the
particle of interest). The number of particles at a given
constant log(dp ) bin was counted using a simple
sorting program to draw a number size distribution as
a function of projected area equivalent diameter.
Figure 8 shows the resulting size distributions for
particles in Figs. 6 and 7 using the data analysis procedure described. Clearly the size distribution has
both shifted to lower sizes and broadened. The results
are qualitatively remarkably consistent with our prior
tandem-DMA oxidation studies in which we observed
changes in size resulting in oxidation using a second
DMA to observe size changes [9].
450
H. Jung et al. / Combustion and Flame 136 (2004) 445–456
(a)
(a)
(b)
(b)
Fig. 6. TEM images before and after oxidation for a group
of particles. (a) Size-classified soot particles before oxidation (dm = 50 nm). (b) After oxidation (4 min at 450 ◦ C of
furnace set temperature).
3.2. Projected area equivalent diameter vs mobility
equivalent diameter
In order to quantify these results it is necessary
to understand the relationship between projected area
equivalent diameter (dA ) and mobility equivalent diameter (dm ). DMA selected particle mobility diameters (50, 100, 130, and 150 nm) and we obtained TEM
samples for each selected size using the procedure
illustrated in Fig. 1a. Both dA and dm are orientationaveraged properties, since the agglomerates are ex-
Fig. 7. TEM images before and after oxidation for a
group of particles after digital filtering to remove background. (a) Size-classified soot particles before oxidation
(dm = 50 nm). (b) After oxidation (4 min at 450 ◦ C of furnace set temperature).
pected to rotate randomly in a DMA column and in
an impactor. Figures 9a, 10, and 11 show size distributions of the TEM samples for a given mobility
diameter as a function of dA . Figure 9a shows that
the selected mobility size from the DMA (50 nm) coincides with the peak of the projected area equivalent
diameter obtained by analyzing the particles on the
TEM grid. We also observe a second peak at 75 nm.
The DMA selects particles of a specified mobility and
while most particles are singly charged a smaller fraction will be doubly charged and will also exit the
H. Jung et al. / Combustion and Flame 136 (2004) 445–456
451
(a)
Fig. 8. Size distribution as a function of projected area equivalent diameter (dA ) before and after oxidation. Data for the
lognormal fit were selected in a monotonic decreasing fashion from the peak for Figs. 8–14. Data that do not obey this
monotonic decrease were excluded for the lognormal fit.
DMA. In this case, the second peak (at 75 nm) corresponds to particles of the same electrical mobility
but with two charges.
This observation is not new. Wen and Kasper [15]
observed a peak from doubly charged particles when
they deposited size-selected chain agglomerates with
a DMA on a TEM grid using a thermal precipitator.
Rogak et al. [16] conducted a very thorough investigation by comparing dA determined by TEM image
analysis with dm determined by a DMA. They observed that dm and dA were essentially equivalent
for TiO2 agglomerates of size less than 400 nm. Figure 9b shows a mobility size distribution measured
using the tandem-DMA method [9] shown in Fig. 1c,
with the furnace turned off. The peak, which is around
75 nm in dm as shown in Fig. 9b, represents particles which were doubly charged when they passed
through the DMA-1 but singly charged when they
passed through the DMA-2 shown in Fig. 1c. The
excellent agreement between Figs. 9a and 9b confirms that dm is effectively equivalent to dA in the
size ranges of interest to this study. Naturally, the
TEM data are noisier because we limited the analysis to only 100 particles, while the DMA effectively
is sampling several orders of magnitude of particles.
Figures 10 and 11 show examples of the number size
distribution as a function of dA for different DMA
size selections (dm = 100 and 150 nm). There was no
obvious peak in the number size distribution found
for multiply charged particles, but there are significant counts at the approximate place where a particle
of equivalent mobility and doubly charged should appear (151 and 234 nm, respectively, for a 100- and
150-nm DMA selection). One likely explanation is
that the peak in the number distribution of soot particles from the flame source is ∼100 nm. This means
the number concentration of larger doubly charged
(b)
Fig. 9. Comparison between size distributions measured as
a function of projected area equivalent diameter (dA ) and
as a function of mobility equivalent diameter (dm ) for a
given mobility diameter selection (dm = 50 nm) by a DMA.
(a) TEM measurement. (b) Tandem-DMA measurement.
Fig. 10. Size distribution as a function of dA when DMA
selects dm = 100 nm.
particles could be unnoticeable at a mobility size selection larger than 100 nm. Figures 12a–12c show
TEM images at the same magnification for dm = 50,
100, and 150 nm, respectively. Figure 13 shows that
the relationship between dm and dA is linear, with a
452
H. Jung et al. / Combustion and Flame 136 (2004) 445–456
Fig. 11. Size distribution as a function of dA when DMA
selects dm = 150 nm.
slope of unity for all measurements including those
shown in Figs. 9a, 10, and 11.
(a)
3.3. Determination of the oxidation rate
The decrease in the mobility diameter of soot particles after undergoing oxidation was translated into
an oxidation rate as in our previous work by Higgins
et al. [9]. The oxidation rate of soot can also be obtained from the decrease in dA by substituting dm
with dA in Higgins et al.’s [9] Eq. (6), since we determined that dA equals dm .
We define the oxidation rate as
dms
(1)
= −ẇAs ,
dt
where ms is the soot particle mass, ẇ the surface specific rate, and As the soot surface area. The surface
area for mass transfer of the soot particles in the free
2 . By definition the effective
molecular regime is πdm
density is
(b)
3
πdm
(2)
.
6
After mass and area are expressed in terms of the mobility equivalent diameter, Eq. (1) becomes
ms = ρeffective
2ẇ
ddm
.
=−
dt
ρeffective
(3)
Since this study does not measure the effective density, small-size agglomerates (dm = 50 nm) were
sampled for the oxidation experiment and assumed
to be of constant density. This is equivalent to saying
they are spherical. The density of soot was assumed to
be 1800 kg/m3 , as suggested by Kennedy [17]. The
mass-based surface specific oxidation rate, ẇ, is often
modeled using a modified Arrhenius expression,
−Ea
,
ẇ = AT 1/2 exp
(4)
RT
(c)
Fig. 12. TEM images at different dm selections. (a) dm =
50 nm, (b) dm = 100 nm, (c) dm = 150 nm.
H. Jung et al. / Combustion and Flame 136 (2004) 445–456
Fig. 13. The relationship between dA and dm for soot agglomerates. dm /2λ is the normalized particle diameter by
2λ, where λ is the mean free path of the carrier gas (here, it
is air at room temperature).
453
(a)
where A is the preexponential factor, Ea is the activation energy, R is the gas constant, and T is temperature in Kelvin.
Since the given temperature was not constant with
time as shown in Fig. 4, Eq. (3) is integrated over exposure time as follows:
t
dm = dA =
−
2ẇ(t)
dt.
ρ
(5)
(b)
0
Equation (5) gives the relationship between the surface specific oxidation rate and the change in projected area equivalent diameter. Given the experimental data of dA at different furnace set temperatures,
a least-square fit was carried out along with the integration shown in Eq. (5) to determine A and Ea for ẇ.
3.4. Comparison of rates with prior studies
In order to help in the validation of this work we
have compared our results with our prior tandemDMA measurements of Higgins et al. [18]. Unfortunately the oxidation experiment of Higgins et al.
[18] could not be performed at the low temperature
range used in this study, because of the long residence
time needed. Instead the comparison was conducted
at the higher temperature range. The comparison then
was between an experiment done with configuration
shown in Figs. 1b and 1c. The size distributions at
25 ◦ C (room temperature) and a furnace temperature
of 1000 ◦ C were compared using DMA-2 and TEM
image analysis and are presented in Fig. 14. The two
experimental methods show excellent agreement in
the most probable particle diameter before and after
oxidation and are further proof that the project area
equivalent diameter, dA , is equivalent to the electrical
mobility diameter, dm .
Fig. 14. Comparison of the results of Higgins et al. [18]
with the current study. (a) TEM measurements before and
after oxidation at 1000 ◦ C. (b) Tandem-DMA measurement
before and after oxidation at 1000 ◦ C of furnace set temperature.
Figure 15 shows Arrhenius plots of the surface
specific oxidation rate for our measurements and
prior studies. The activation energy of low-temperature flame soot oxidation was determined to be
148 kJ/mol for low temperatures and is consistent
with the 143 kJ/mol used in the Nagle and StricklandConstable (NSC) [19] oxidation model (curve d in
Fig. 15) as extracted by Stanmore et al. [20]. However, our oxidation rate is an order of magnitude
higher than the extrapolated NSC model and even
more than that for the other studies presented in this
temperature range. To assess the possibility that this
experimental approach somehow results in a higher
observed reaction rate, we conducted an experiment
under the same conditions (i.e., higher temperature)
as in our prior tandem-DMA oxidation work [9,18].
Here the setup shown in Fig. 1b was used and the
TEM was used to find size distributions before and
after the furnace. The result is shown in Fig. 15 as
the open circle, and overlaps with excellent agreement with rate curve c, of Higgins et al. [18]. This
454
H. Jung et al. / Combustion and Flame 136 (2004) 445–456
Fig. 15. Arrhenius plot of the surface specific rates of soot oxidation for the current study and for other relevant prior studies.
Previous studies produced a variety of activation energies, as can be seen by the varying slopes of the lines. Results are presented
for 21% O2 . (Curve a) Current study (—), rates for ±1σ confidence lines (- - -), rates for ±20 K variation (—). (Curve b) Higgins
et al. [9] using flame-generated soot. (Curve c) Higgins et al. using diesel soot [18]. (Curve d) Nagle and Strickland-Constable
[19] using a pyrographite rod. (Curve e) Miyamoto et al. [23] TGA of catalyzed (Ca added to fuel) diesel soot. (Curve f)
Miyamoto et al. [23] TGA of uncatalyzed diesel soot. (Curve g) Ahlström and Odenbrand [3] flow reactor study of diesel soot.
(Curve h) Otto et al. [1] TGA of diesel soot. (Curve i) Gilot et al. [2] TGA of carbon black. (Curve j) Neeft et al. [4] flow reactor
study of Printex-U flame soot from Degussa AG.
would suggest that the TEM sizing approach should
be yielding accurate results at the lower temperature
conditions.
To assess other possibilities for our higher rate,
the soot sampled on the TEM grid was pretreated under a N2 environment at 550 ◦ C for 5 min to rule out
the possibility that volatile organics may be enhancing the apparent oxidation rate. The temperature and
residence time for the pretreatment were taken from
the IMPROVE method [21]. We observed no noticeable size change after N2 pretreatment, nor was there
any subsequent change in the observed oxidation rate.
Another possibility considered was that the soot
which was in contact with the SiO2 TEM grid might
somehow accelerate oxidation; however, van Doorn
et al. [22] have shown that SiO2 has no catalytic effect on diesel soot particles. One may expect a lower
activation energy along with an increased rate if catalytic effects are important as observed by Miyamoto
et al. [23]. In their work they saw enhanced rates for
calcium-laden diesel soot of up to 2 orders of magnitude higher (curve e in Fig. 15) compared with undoped soot (curve f in Fig. 15).
Temperature is one of the most sensitive variables
in determining reaction rate. Our uncertainty analysis
discussed in Appendix A indicates an uncertainty of
less than ±2 K. If one were to assume a worst case
scenario and assume an uncertainty 10 times larger,
the resulting effect on the rate is shown in Fig. 15,
as the thin lines above and below rate curve a. Obviously uncertainty in temperature cannot explain our
observed higher rate.
The uncertainty related to the image analysis is
also discussed in Appendix A. The uncertainty in the
measurement of a 20-nm diameter is expected to be
less than ±2 nm. However, for particles below 20 nm,
the uncertainty increases due to a decrease in the number of pixels which constitute a given particle. The
dotted lines seen in Fig. 15, curve a, show ±1σ confidence lines. The propagation of the uncertainty in
determination of dA to the final rate is negligible,
since it lies within the confidence line limits above.
Stanmore et al. [20] in their recent review have discussed the variation in the measured oxidation rates,
some of which are presented in Fig. 15. They suggested the possibility that some of the measured rates
might be corrupted by mass transfer effects, which
would tend to make the observed rate lower. We
should also note that in the low temperature range,
curve j reflects results for black carbon, while curves
H. Jung et al. / Combustion and Flame 136 (2004) 445–456
h and g are diesel soot results from Otto et al. [1]
and Ahlström and Odenbrand [3], respectively. This
difference in source may also create differences in
measured rates.
Ishiguro et al. [6] studied the microstructure evolution for diesel soot oxidation at about 800 K using
an analytical electron microscope. During oxidation
they observed flaking of crystallites from the outermost shell of the particles. This would certainly make
the apparent rate higher than NSC prediction; however, that does not explain the discrepancy between
our results and the other low-temperature data. Furthermore, given the resolution of our instruments we
were not able to assess if this flaking phenomenon
was taking place for our particles.
So at this point we are left without a firm conclusion as to why we observe these differences with prior
results. Since the imaging method presented here gave
results similar to those of our tandem-DMA studies under the same conditions, we conclude that our
higher rate is attributable to a measurement protocol
that more effectively assesses the intrinsic chemical
reactivity, without the corruption of heat and mass
transfer effects associated with bulk methods. These
results are consistent with our single-particle massspectrometry studies on condensed-phase chemical
kinetics, in which we observed much higher reaction
rates for small particles over bulk samples [5].
4. Conclusion
An experimental protocol was developed to visualize oxidation of soot at low temperatures and to
measure the oxidation rate of soot particles using
TEM image analysis. The visualization of individual soot particles showed the disappearance of a portion of the soot aggregates during oxidation. A group
of soot particles was tracked to better represent the
changes in dA during oxidation. TEM samples using an impactor after size classification by DMA gave
number distribution as a function of projected area
equivalent diameter, dA , at given mobility diameter, dm . The size distributions obtained from the TEM
images confirm that dm equals dA as Rogak et al. [16]
observed using TiO2 agglomerates. The relationship
between dm and dA was used to convert the change
in projected area to the oxidation rate of soot particles
after slightly modifying the conversion equation developed by Higgins et al. [9]. The surface area specific
oxidation rate of soot particles at temperatures below
800 K was found higher than that found in prior studies. The rates determined using the TGA method may
be lower due to limitations of heat and mass transfer
that Mahadevan et al. [5] found in the bulk method.
In conclusion, the experimental protocol developed
455
for this project can be used to measure reaction rates
and to visualize the reaction at the same time for a
variety of applications, overcoming the limitation of
heat and mass transfer seen in the conventional thermal method. Further work is necessary to identify
what caused the difference between low- and hightemperature oxidation rates.
Acknowledgments
The authors are grateful for the help received from
Dr. Stuart McKernan and Dr. David Bell concerning
TEM. The authors also thank Professor Peter McMurry and Dr. Kihong Park for allowing us to use
their sampling-impactor system.
Appendix A. Uncertainty and error analysis
A.1. Uncertainty related to temperature
measurement
Temperature measurement is very important in determining reaction rates since temperature is one of
the most sensitive variables. The gas temperature profile shown in Fig. 4 is based on temperature measurements taken on top of the TEM grid after the crucible
was placed in the middle of the furnace as shown in
Fig. 3. The K-type thermocouple, with a diameter of
0.05 mm (Model CHAL 001; Omega), was used and
its response time is approximately 0.05 s due to its
ultralow heat capacity. To confirm whether the temperature of a TEM grid follows the gas temperature,
temperature readings were taken using the same type
of thermocouple as that attached to the TEM grid.
The increase in total heat capacity due to the attached
TEM grid did not cause noticeable temperature delay,
which implies that the TEM grid also has a negligible amount of heat capacity. The effect of radiation
heat transfer was assessed by measuring steady-state
temperatures using two types of thermocouple probes
after 4 min of elapsed time. A radiation-shielded thermocouple probe (Omega) and an exposed-tip thermocouple probe (Omega) with a tip diameter of 1/16 in.
were used to measure the gas temperature after 4 min
of elapsed time. These two probes have longer response times due to their high heat capacity. The
above four measurements showed agreement within
±1.5 K, after 4 min of elapsed time, and implies that
radiation heat transfer effects could be ignored.
A.2. Uncertainty related to the TEM image analysis
Determining a threshold for digital filtering is the
most important procedure for differentiating soot particles from background. In particular, images with
456
H. Jung et al. / Combustion and Flame 136 (2004) 445–456
of less than 2 nm for a 20-nm-diameter primary particle.
References
(a)
(b)
Fig. 16. Typical histograms of the TEM images. (a) Ideal
case. (b) As observed (reproduced, by permission of the publisher, from [24]).
high contrast between particles and background are
best for data analysis. Figure 16 shows histograms of
pixel counts as a function of color brightness. Ideally,
a histogram of an image containing particles will have
two major peaks as shown in Fig. 16a: one for the
background and the other for the particles. In practice, the peaks may not be clearly distinguishable as
shown in Fig. 16b (Gatan, Inc. [24]).
Because each TEM image will have a slightly
different contrast determination of the background,
threshold level varies from image to image. It was
necessary therefore to manually set the threshold for
each image. Several images were taken for the same
field of view to determine uncertainty in determining dA . Each image was thresholded with five different settings and the average diameter and upper
and lower limit diameters of a given particle were determined from each thresholding of all the images.
The results of this analysis indicated an uncertainty
[1] K. Otto, M.H. Sieg, M. Zinbo, L. Bartosiewicz, SAE
Technol. Pap. 800336, 1980.
[2] P. Gilot, F. Bonnefoy, F. Marcuccilli, G. Prado, Combust. Flame 95 (1993) 87.
[3] A.F. Ahlström, C.U.I. Odenbrand, Carbon 3 (1989)
475.
[4] J.P.A. Neeft, T.X. Nijhuis, E. Smakman, M. Makkee,
J.A. Moulijn, Fuel 76 (1997) 1129.
[5] R. Mahadevan, D. Lee, H. Sakurai, M.R. Zachariah,
J. Phys. Chem. A 106 (2002) 11083–11092.
[6] T. Ishiguro, N. Suzuki, Y. Fujitani, H. Morimoto, Combust. Flame 85 (1991) 1.
[7] T. Tanaka, in: 22nd International Vienna Engine Symposium, vol. 2, OVK and Technical Univ. of Vienna,
2001.
[8] R.J. Santoro, H.G. Semerijan, R.A. Dobbins, Combust.
Flame 51 (1983) 203.
[9] K.J. Higgins, H. Jung, D.B. Kittelson, J.T. Roberts,
M.R. Zachariah, J. Phys. Chem. A 106 (2002) 96–103.
[10] S.V. Hering, S.K. Friedlander, J.J. Collins, L.W.
Richard, Environ. Sci. Technol. 13 (1979) 184–188.
[11] F.A. Heckman, Rubber Chem. Technol. 37 (1964)
1245.
[12] P. Gilot, F. Marcuccilli, G. Prado, Corrosion of Advanced Ceramics, 1963, pp. 329–339.
[13] W. Hess, M.C.R. Herd, in: J.B. Donnet, R.C. Bansal,
M.-J. Wang (Eds.), Carbon Black, second ed., Dekker,
New York, 1993, chapter 3.
[14] NIH Image, Image software developed at the U.S. National Institutes of Health and available on the Internet
at http://rsb.info.nih.gov/nih-image/.
[15] H.Y. Wen, G. Kasper, Aerosol Sci. Technol. 3 (1984)
397.
[16] S.N. Rogak, R.C. Flagan, H.V. Nguyen, Aerosol Sci.
Technol. 18 (1993) 25.
[17] I.M. Kennedy, Prog. Energy Combust. Sci. 23 (1997)
95.
[18] K.J. Higgins, H. Jung, D.B. Kittelson, J.T. Roberts,
M.R. Zachariah, Environ. Sci. Technol. 37 (2003)
1949–1954.
[19] J. Nagle, R.F. Strickland-Constable, in: Fifth Carbon
Conference, vol. 1, Pergamon, Oxford, 1962, pp. 154–
164.
[20] B.R. Stanmore, J.F. Brilhac, P. Gilot, Carbon 39 (2001)
2247–2268.
[21] J.C. Chow, J.G. Watson, D. Crow, D.H. Lowenthal,
T. Merrifield, Aerosol Sci. Technol. 34 (2001) 23.
[22] J. van Doorn, J. Varloud, P. Mérıaudeau, V. Perrıchonı,
Appl. Catal. B 1 (1992) 117.
[23] N. Miyamoto, H. Zhixin, O. Hideyuki, SAE Pap.
881224, 1988.
[24] Gatan, Inc., Digital Micrograph User’s Manual v.3.3,
1998.